What it genuinely does well
Reading a messy, unordered dump and returning a sensible ranking is squarely within what language models are good at. It handles fragments, mixed contexts and implied deadlines without you formatting anything.
It's also consistent. It doesn't rank differently because it's Friday afternoon, and it has no emotional stake in avoiding the difficult item — which is a genuine advantage over ranking your own list at 8am.
What it cannot do
It doesn't know what you haven't told it. If a low-stakes-looking task is actually critical because of something you didn't write down, it will rank it low.
It cannot make the work smaller, and it cannot decide what you value. Those are yours.
On the claims in this category
You will see specific productivity percentages advertised across this category. Ask which study produced the figure, who ran it, and against what control group. Usually there is no study at all — the number was written by a marketing team.
We don't publish a figure for Lumin because we haven't run a controlled trial. Neither, as far as we can establish, has anyone else in this category.
What the underlying research supports
The well-evidenced part isn't the AI — it's externalisation. Masicampo & Baumeister (2011) showed that a specific plan removes the intrusive pull of unfinished goals, and Risko & Gilbert (2016) reviews how external tools reduce working memory demand.
AI's contribution is doing the ranking step quickly and consistently. That's a real convenience. It isn't a new psychological mechanism.
Related guides
Frequently asked questions
Can AI know what's important to me?
Only from what you write. It infers from deadlines, stated consequences and context in the dump. Unstated context is invisible to it, which is why the ranking is a starting point you can override.
Is AI prioritisation better than doing it myself?
It's faster and more consistent, and it doesn't flinch from the task you're avoiding. Your own judgement is better informed. In practice the useful combination is letting it produce the order and correcting it where you know better.
What if the ranking is wrong?
Change it. A ranking you disagree with has still done something useful — it's given you something concrete to react to, which is much easier than generating an order from scratch.